Professional Documents
Culture Documents
All correspondence concerning this article should be addressed to Martin Pinquart, Department of
Psychology, Philipps University, D-35032 Marburg, Germany. E-mail: pinquart@staff.uni-marburg.de
Received December 29, 2010; revisions received May 3, 2011; accepted May 15, 2011
Objective To examine the risk of emotional and behavioral problems among children with a chronic physi-
cal illness. Methods Random-effects meta-analysis was computed to integrate the results of 569 studies
that used the Child Behavior Checklist, Youth Self Report, and the Teacher Report Form. Results Young
people with a chronic physical illness have higher levels of internalizing (g ¼ .47 standard mean difference),
externalizing (g ¼ .22) and total behavior problems (g ¼ .42) than healthy peers. The largest differences
were found in parental ratings and the weakest differences in adolescent self-ratings. Strongest elevations of
internalizing problems were found for chronic fatigue syndrome and strongest elevations of externalizing
problems were observed for epilepsy and migraine/tension-type headache. Effects also varied by country and,
in part, by age, gender, year of publication, and study design. Conclusions The results call for regular
screens for psychological distress and referrals for mental health services, when needed.
Key words Achenbach System of Empirically Based Assessment; Child Behavior Checklist; chronic illness;
externalizing problems; internalizing problems; meta-analysis.
Epidemiological studies have shown that about 15% of chil- problems include delinquent and aggressive behaviors, and
dren and adolescents, on average, have a chronic physical internalizing problems include somatic complaints, anxi-
health condition (van der Lee, Mokkink, Grootenhuis, ety, depression, and social withdrawal.
Heymans, & Offringa, 2007). These conditions are risk fac- Elevated levels of internalizing as well as externaliz-
tors for behavior problems (e.g., Barlow & Ellard, 2006). ing problems may be observed in children with chronic
Analyzing the prevalence of behavior problems of children physical illness (Barlow & Ellard, 2006; Lavigne & Faier-
with chronic physical illness and identifying related risk Routman, 1992). Sources of elevated internalizing prob-
factors is of theoretical and practical relevance. First, it lems in children with a chronic physical illness may be
helps to understand psychosocial consequences of chronic the perceived lack of control over the illness and its symp-
physical diseases. Second, it provides valuable information toms or progression (e.g., in epilepsy or sickle cell disease),
for clinicians regarding who should be screened for which frightening symptoms (e.g., in the case of asthma or
kind of problems. Third, it provides important information epilepsy), restrictions in positive activities (e.g., due to hos-
about the need for preventing these problems as part of an pitalization), peer rejection (e.g., in the case of visible ab-
integrated treatment. Since many chronic diseases cannot normalities, such as cleft lip and palate), as well as side
be cured, a main goal is to reduce the consequences of the effects of therapy (e.g., in the case of radiation and chemo-
diseases on the lives of children and adolescents. therapy of cancer patients). In addition, symptoms of the
Multivariate approaches to child psychopathology physical illness, such as pain, may lead to elevated scores
often distinguish between externalizing and internalizing of somatic complaints that are part of internalizing problem
problems and disorders (Achenbach, 1991). Externalizing clusters (Achenbach, 1991).
If behavior problems are gender-specific responses to [Child Behavior Checklist or Youth Self Report or Teacher
illness-related stressors, these gender differences may be Report Form or Achenbach System of Empirically Based
accentuated in young people with chronic illnesses. Assessment]) and cross-referencing. Criteria for inclusion
of studies in the present meta-analysis were:
Country
(a) they used the CBCL (CBCL 4–18, CBCL 1½–5),
On average, better medical and psychosocial services may
YSR or TRF;
be available in industrial countries than in developing or
(b) the studies were published since the development
threshold countries. Therefore, we expected to find smaller
of the scales and before May, 2011;
effect sizes in studies from industrialized countries.
(c) they compared levels of behavior problems between
Publication Status young people with chronic physical illnesses and
healthy peers or test norms, or they provided suffi-
As nonsignificant results may be less likely to be published
cient information for a comparison with established
(Lipsey & Wilson, 2001), we wanted to test whether
normative data (e.g., by reporting standardized
unpublished studies (e.g., dissertations) would report
T-scores, raw score, and percentage of respondents
lower effect sizes than published studies.
above defined cut-offs);
Year of Publication (d) mean age of participants was 18 years; and
(e) standardized between-group differences in behavior
We expected to find lower between-group differences in
problems were reported or could be computed.
more recent studies because of progress in the treatment of
many diseases (e.g., Bleyer, 2002) and the development of Documentation of a physician’s diagnosis within each
psychosocial services for young people with chronic illness. study was not a requirement because of the need to include
broad-based survey studies for which medical documenta-
Study Quality tion might not have been available. In order to include
Some clinical samples may over-represent highly distressed studies from different regions of the world, we also did
young people seeking treatment for their chronic disease, not limit the included studies to those written in English.
which is why we expected stronger between-group differ- Studies published in Chinese, Dutch, English, German,
ences in clinical (convenience) samples than in random Korean, Portuguese, Spanish, and Turkish were included.
community-based samples. Available unpublished studies (e.g., dissertations) were
also included. Studies were excluded if the samples of chil-
Exploratory Analysis of Other Moderators dren were identified because they had ‘‘clinical’’ levels of
Moderating effects of duration of illness, race/ethnicity, behavior problems.
target of comparison (test norms vs. healthy control group), We identified 693 studies. After screening and assess-
and sociodemographic matching of the patient and control ing for eligibility, we were able to include 569 studies in
group were assessed exploratively, because no theory-based the meta-analysis. A flow chart of the search for studies is
expectations could be stated and/or because results from provided in Appendix A1, and the studies included are
previous meta-analyses are contradictory. listed in the Appendix A2 (see Supplementary Data).
Most of the studies were done in industrialized countries.
However, we were able to include studies from Brazil,
Methods China, the Dominican Republic, India, Egypt, Malaysia,
Sample Thailand, and Turkey. Information from the World Bank
Studies were identified through electronic databases (2010) was used for coding countries as developed or de-
(PSYCINFO, MEDLINE, Google Scholar, and PSYNDEX veloping/threshold countries.
[an electronic data base of psychological literature from We entered the number of patients and control group
German-speaking countries] search terms: [chronic illness members, mean age, percentage of girls and percentage of
or disability or AIDS or arthritis or asthma or cancer or members of ethnic minorities, country of data collection,
chronic fatigue syndrome or cleft or cystic fibrosis or deaf year of publication, type of illness, duration of illness, the
or diabetes or epilepsy or headache or heart disease or sampling procedure (1 ¼ probability samples, 0 ¼ conve-
hearing impairment or HIV or inflammatory bowel disease nience samples), use of a control group (0 ¼ yes, 1 ¼ com-
or kidney disease or liver disease or migraine or rheuma- parison with test norms), equivalence of patient and control
tismor sickle cell or spina bifida or visual impairment] and group (1 ¼ yes, 2 ¼ not tested, 3 ¼ no), the rater of behavior
1006 Pinquart and Shen
problems (1 ¼ adolescent, 2 ¼ parent, 3 ¼ teacher), and 4. Homogeneity of effect sizes was computed by use
the standardized size of between-group differences in of the Q-statistic.
total problems, internalizing problems, externalizing prob- 5. In order to test the influence of moderator
lems, withdrawn behavior, somatic complaints, anxiety/ variables, we used an analogue of an analysis of
depression, delinquent behavior, aggressive behavior, social variance and weighted ordinary least squares
problems, thought problems, and attention problems. If regression analyses. In univariate analyses, a signif-
between-group differences were provided for several sub- icant Qb-score indicates heterogeneity of the effect
groups within the same publication (e.g., for different ill- sizes between the compared conditions. Which
nesses), we entered them separately in our analysis instead conditions differ is tested by comparing the 95%
of entering the global association. If data from more than CIs. Differences between two conditions are signif-
one rater were collected, we entered the effect sizes sepa- icant if CI of two effect sizes do not overlap
rately. All studies were coded by the first author, and one- (Lipsey & Wilson, 2001). Statistical power analysis
third of them also by the second author. A mean inter-rater was computed based on Hedges and Pigott
reliability of 94% (range 89–100%) was established. (2001).
Differences were resolved by discussion.
Results
Data from 51,422 children and adolescents with chronic
Statistical Integration of the Findings
illnesses were included. The largest subgroup had asthma
Calculations for the meta-analysis were performed in five (n ¼ 13,793), followed by epilepsy (n ¼ 6,815), cancer
steps, using random-effects models and the method of (n ¼ 3,936), heart diseases (n ¼ 2,692), diabetes (n ¼ 2,136),
moments (Lipsey & Wilson, 2001). hearing impairments (n ¼ 2,136), sickle cell disease
(n ¼ 1,897), cleft lip and palate (n ¼ 1,817), arthritis
1. We computed effect sizes d for each study as the (n ¼ 1,746), migraine/tension headache (n ¼ 1,188),
difference in behavior problems between the kidney/liver disease (n ¼ 937), HIV-infection (n ¼ 771),
sample with chronic illnesses and the control spina bifida (n ¼ 700), cystic fibrosis (n ¼ 674), inflamma-
sample divided by the pooled SD. If the authors tory bowel disease (n ¼ 579), chronic fatigue syndrome
only provided test scores for children and adoles- (n ¼ 289), and visual impairment (n ¼ 186). The partici-
cents with chronic illness, we used the norms pants had a mean age of 10.6 years (SD ¼ 6.8 years);
from the test manuals for comparison. If separate 45.9% of them were girls and 43.7% were members of
CBCL scores were reported for mothers and ethnic minorities.
fathers, we averaged both scores. Outliers that Power analysis indicates that the probability was
were more than 2 SD from the mean of the effect .8 that the study will detect a small effect sizes of d ¼ .20.
sizes were recoded to the value at 2 SD, based on
Lipsey and Wilson (2001). Levels of Internalizing, Externalizing, and Total
2. Using Hedges’ g, effect size estimates were adjusted Problems
for bias due to overestimation of the population When summing up all effect sizes, we observe higher levels
effect size in small samples. Hedges’ g leads of total problem behavior in children with chronic illness
to smaller estimations of the effect size than as compared to healthy peers or test norms (g ¼ .42, 95%
Cohen’s d, but the difference declines with CI 0.38–0.45, Z ¼ 24.93, p < .001). Similarly, levels of in-
increasing sample size and is less than 1% for ternalizing problems (g ¼ .47, CI 0.44–0.50, Z ¼ 29.73,
sample sizes of 80 and larger. p < .001) and externalizing problems (g ¼ .22, CI .19–
3. Weighted mean effect sizes and 95% confidence .25, Z ¼ 14.76, p < .001) were elevated in children with
intervals (CIs) were computed. The significance chronic illness.
of the mean was tested by dividing the weighted According to Cohen’s criteria for interpreting effect
mean effect size by the standard error of the sizes (Cohen, 1988), effects were small to moderate
mean. To interpret the practical significance of the when using parental and teacher ratings. In studies using
results, we used Cohen’s criteria (Cohen, 1988): adolescent ratings, effects on total problems and internal-
Differences of g ¼ 0.20–0.49 are small, of izing problems were small, and no significant effect size
g ¼ 0.50–0.79 medium, and of g 0.8 large. was found for externalizing problems (Table I).
Behavior Problems and Chronic Illness 1007
720
953
588
0
0
0
661
376
593
45,447
1,037
As shown by the Qb -statistics, the effect sizes of total
k0
problems differed between raters (parent rating: gP ¼ .46,
teacher rating: gT ¼ .37, child/adolescent rating: gC ¼ .17;
66.03*
64.86
48.68
40.26
82.59
109.38
55.09
55.40
41.72
56.01
29.97
Qw Table I). Effect sizes for internalizing problems
(gP ¼ .51, gT ¼ .38, gC ¼ .27) and externalizing problems
Adolescent reports
.08
.26
.22
.27
.36
.45
.67
.37
.39
.32
.39
95%-CI
Note. k ¼ number of studies; g ¼ effect size; 95% CI ¼ lower and upper limits of 95% CI; Q ¼ test for homogeneity of effect sizes between conditions (Qb) and within conditions (Qw); k0 ¼ fail-safe N.
.10
.06
.02
.07
.18
.22
.38
.14
.12
.09
.11
problems (gP ¼ .24, gC ¼ .01) as well as total problems
(gP ¼ .46, gC ¼ .17) in children with chronic illnesses
.17***
.27***
.34***
.53***
.25***
.26***
.21***
.25***
were more pronounced in studies that used parental re-
g
.01
.10
.10
ports than in studies using child self-reports. In addition,
the non-overlap of the CI indicates that effect sizes for total
73
88
46
47
46
72
41
44
45
41
43
k
184
490
588
2,470
5,724
1,691
1,329
1,291
k0
20.52
20.42
19.60
Qw
.86
.76
.90
95%-CI
.49
.40
.52
.67***
.58***
.71***
25
21
26
k
Table I. Behavior Problems in Children With Chronic Illness: Variations by Source of Information
78,944
57,668
125,992
k0
237.35
210.72
266.57
Qw
ences were only significant for parent ratings (gP ¼ .51 vs.
.49
.54
.62
.78
.57
.27
.39
.45
.66
.53
.72
.54
.43
.60
.60***
.48***
.66***
215
221
244
k
29.69***
26.51***
19.132**
28.57***
20.19***
22.43***
19.32***
27.18***
12.76**
11.50**
between
7.20*
Attention problems
Anxious/depressed
Thought problems
Social problems
Delinquent
Aggressive
subscale, we computed a modified total problems score of young people with epilepsy and kidney/liver diseases,
the CBCL by averaging the effect sizes of all subscales levels of internalizing problems in children with arthritis,
except somatic complaints. Similarly, we computed a mod- heart diseases, and kidney/liver diseases, as well as levels of
ified Internalizing Problems score by averaging the effect externalizing problems of young patients with chronic fa-
size of the withdrawn behavior and anxiety/depression. tigue syndrome, kidney/liver diseases, and the sum cate-
This procedure was possible for about 50% of the available gory of other illnesses.
studies. As indicated in the Supplementary Appendix 3, Scores on the CBCL subscales for the individual dis-
the modified scales showed moderate effects for total prob- eases are reported in Table III. Effect sizes for all subscales
lems (gP ¼ .51) and internalizing problems (gP ¼ .53), and varied between disease groups. According to Cohen’s cri-
their CI overlapped with the CIs of the original scales in teria (1988), large effect sizes were found for spina bifida
Table I. Thus, the results were quite similar for the original (attention problems: gP ¼ 1.00, social problems: gP ¼ .94,
and modified scales. withdrawal: gP ¼ .93), epilepsy (attention problems: gP ¼
1.07, social problems: gP ¼ .88), chronic fatigue syndrome
Comparison of Illnesses (anxious-depressiveness: gP ¼ .99, somatic complaints:
The next analysis focused on the comparison of the differ- gP ¼ 1.35), HIV-infection/AIDS (withdrawal: gP ¼ 1.44, de-
ent kinds of illnesses. Separate effect sizes were reported if linquency: gP ¼ .84), inflammatory bowel disease (with-
at least five studies were available for a particular illness. drawal: gP ¼ 84, somatic complaints, gP ¼ 1.47), kidney/
Since studies on vision loss and hearing loss did not report liver disease (somatic complaints: gP ¼ .88, attention prob-
separate effect sizes for the associated different diseases, we lems: gP ¼ .86), asthma (somatic complaints: gP ¼ 1.02),
used only sum categories of vision and hearing loss. The arthritis/rheumatism (somatic complaints: gP ¼ .92), and
following analyses are based on parental reports of their sickle cell disease (somatic complaints: gP ¼ .84).
children’s behavior.
As indicated by the Q-statistics (Table II, first line), the Moderator Effects of Sociodemographic
effect sizes for internalizing, externalizing, and total prob- Characteristics
lems varied between diseases. Statistically significantly ele- In order to compare our results on age differences with the
vated levels of total behavior problems were observed for all study by Karsdorp et al. (2007), we divided the sample into
assessed diseases except for cleft lip and palate and HIV the same age groups. The Q-statistics indicates significant
infection. Elevated levels of internalizing problems were age differences for the three problem scales. The non-
found for all diseases except cleft lip and palate. With overlap of the CI indicates that effect sizes were smaller
regard to total problems, moderate effect sizes were found in children under the age of 6 years than in the other age
for migraine/tension-type headache (gP ¼ .75), kidney/liver groups for total problems (gP ¼ .19 vs. gP ¼ .48), internal-
disease (gP ¼ .70), epilepsy (gP ¼ .63), chronic fatigue izing (gP ¼ .18 vs. gP ¼ .48–.55), and externalizing prob-
syndrome (gP ¼ .62), asthma (gP ¼ .61), hearing impair- lems (gP ¼ .06 vs. gP ¼ .25–.28; Table II). The effect sizes
ment (gP ¼ .56), and spina bifida (gP ¼ .50). According to within each age group were homogeneous, except the total
Cohen’s (1988) criteria, elevations of internalizing prob- score and the internalizing score in the youngest group.
lems were large in children with chronic fatigue syndrome In order to explore moderating effects of gender, we
(gP ¼ 1.42). In addition, moderate elevations were observed trisected the sample according to the percentage of female
with regard to migraine/tension headache (gP ¼ .77), epilepsy participants. Significant gender differences were observed
(gP ¼ .66), chronic kidney/liver disease (gP ¼ .67), asthma with regard to total and externalizing problems. The non-
(gP ¼ .63), inflammatory bowel disease (gP ¼ .60), and overlap of the CI indicates that the levels of total (gP ¼ .58)
spina bifida (gP ¼ .59). Small effect sizes are listed in and externalizing problems (gP ¼ .38) showed significantly
Table II. higher elevations if the samples were comprised of
Elevated levels of externalizing problems in children with two-thirds or more male participants.
chronic illness were only observed for 10 out of 17 diseases/ Furthermore, we observed significant differences be-
disabilities (Table II). Interestingly, a negative effect size was tween children from developing and developed countries
observed for externalizing problems of children with cleft for all three problem scales. Stronger elevations of prob-
lip and palate (gP ¼ .24): Parents of these children re- lems were found in children with chronic illnesses living in
ported fewer symptoms than parents of healthy children. developing/threshold countries (gP ¼ .47–.65) than in de-
The effect sizes for most illness groups were homoge- veloped countries (gP ¼ .22–.49). However, no significant
neous. However, heterogeneity beyond sampling error was moderating effects were found for ethnicity and duration of
observed with regard to the levels of total problems in illness.
Table II. Association of Study Characteristics With Elevated Levels of CBCL Behavior Problems in Children With Chronic Illness
Arthritis/rheumatism 10 .56*** .31 .80 11 .92*** .63 1.21 7 .49*** .21 .77 7 .10 .19 .39
Asthma 9 .46*** .20 .72 15 1.02*** .76 1.28 14 .64*** .46 .82 14 .44*** .22 .66
Cancer 25 .50*** .35 .66 30 .77*** .60 .94 20 .47*** .31 .64 23 .13 .03 .29
Chronic fatigue syndrome 9 .77*** .49 1.05 10 1.35*** 1.02 1.68 7 .99*** .67 1.31 8 .03 .32 .25
Kidney/liver disease 18 .76*** .58 .94 17 .88*** .65 1.11 19 .68*** .51 .85 13 .31** .09 .52
Cleft lip and palate 22 .41*** .26 .56 22 .45*** .25 .64 23 .32*** .18 .47 16 .31*** .14 .48
Cystic fibrosis 4 .06 .49 .37 6 .67*** .26 1.09 3 .18 .35 .70 6 .68*** .34 1.03
Diabetes 12 .40*** .20 .61 13 .65*** .40 .91 13 .43*** .24 .63 14 .21* .02 .40
Epilepsy 35 .74*** .63 .86 38 .78*** .63 .92 36 .63*** .51 .74 35 .48*** .37 .59
Hearing impairment 8 .38*** .16 .60 9 .29* .01 .57 8 .36** .14 .59 7 .34** .11 .57
Heart disease 8 .32** .10 .54 7 .40* .09 .71 8 .21 .01 .43 4 .10 .21 .41
HIV infection/AIDS 1 1.44*** .64 2.24 4 .06 .43 .55 3 .62** .17 1.06 2 .84** .27 1.40
IBD 5 .84*** .52 1.15 5 1.47*** 1.06 1.88 4 .60*** .27 .93 3 .20 .17 .57
Kidney/liver disease 18 .76*** .58 .94 17 .88*** .65 1.11 19 .68*** .51 .85 13 .31** .09 .52
Migraine 12 .39*** .20 .59 17 .78*** .56 1.00 10 .51*** .30 .72 11 .07 .13 .27
Sickle cell disease 7 .46** .15 .75 10 .84*** .53 1.14 3 .22 .16 .59 5 .26 .06 .59
Spina bifida 4 .93*** .55 1.31 3 .59* .06 1.12 1 .30 .33 .94 3 .31 .11 .73
Visual impairment 3 .26 .22 .75 5 .28 .19 .75 2 .14 .47 .75 5 .14 .24 .51
Other illnesses 66 .61*** .52 .70 76 .64*** .54 .75 71 .57*** .48 .66 72 .43*** .34 .52
Aggressive Social problems Attention problems Thought problems
(Qb ¼ 51.87***) (Qb ¼ 99.19***) (Qb ¼ 89.33***) (Qb ¼ 55.81***)
Arthritis/rheumatism 6 .45** .12 .77 6 .11 .19 .41 7 .16 .16 .49 8 .33* .06 .60
Asthma 19 .29** .09 .49 10 .05 .19 .29 15 .26* .02 .50 12 .28* .06 .49
Cancer 26 .24** .08 .40 20 .58*** .40 .76 19 .44*** .24 .64 18 .27*** .11 .43
Chronic fatigue syndrome 8 .11 .42 .19 6 .21 .15 .57 5 .54* .13 .95 5 .41* .06 .75
Cleft lip and palate 20 .25** .08 .42 17 .51*** .34 .69 15 .50*** .29 .71 15 .47*** .30 .63
Cystic fibrosis 6 .47* .11 .83 3 .31 .23 .86 4 .74** .25 1.23 6 .33* .02 .65
Diabetes 14 .28** .08 .49 13 .24* .03 .44 13 .41*** .18 .65 14 .23** .06 .40
Epilepsy 39 .62*** .50 .74 35 .88*** .76 1.00 42 1.07*** .94 1.19 38 .74*** .64 .84
Hearing impairment 9 .41*** .18 .64 7 .55*** .28 .81 10 .62*** .38 .86 7 .55*** .44 .77
Heart disease 7 .14 .11 .40 6 .49*** .21 .76 8 .48** .22 .75 3 .41** .10 .73
HIV infection/AIDS 3 .10 .37 .57 3 .28 .16 .72 3 .62* .11 1.13 2 .53* .01 1.05
IBD 4 .42* .04 .80 2 .25 .24 .73 1 .00 .71 .71 2 .43* .04 .83
Kidney/liver disease 18 .70*** .50 .89 14 .64*** .43 .86 16 .86*** .64 1.08 13 .40*** .20 .59
Migraine 11 .27* .05 .49 11 .40*** .19 .61 10 .41*** .16 .66 11 .31*** .13 .49
Sickle cell disease 9 .30* .01 .58 5 .19 .13 .52 6 .33 .08 .74 5 .33* .01 .65
Spina bifida 4 .56** .16 .97 3 .94*** .52 1.36 4 1.00*** .56 1.43 2 .40 .05 .84
Visual impairment 5 .40* .01 .80 2 .42 .23 1.08 5 .53* .09 .97 3 .36 .09 .81
Other illnesses 79 .44*** .35 .52 52 .83*** .72 .93 63 .70*** .60 .81 58 .54*** .46 .63
Note. k ¼ number of studies; g ¼ effect size; 95% CI ¼ lower and upper limits of 95% CI; Qb ¼ test for homogeneity of effect sizes across the different diseases.
*p < .05, **p < .01, ***p < .001.
Moderator Effects of Characteristics of the three decades. The Qb-statistics indicates that the effect
Publication size for externalizing problems, but not for internalizing
The Qb-statistics indicates that published and unpublished and total problems, differ by year of publication. As
studies did not differ in effect size. As the CBCL was in- shown by the nonoverlap of the CI, elevations of external-
troduced 30 years ago, we compared results from these izing problems of children with chronic illness were
1012 Pinquart and Shen
stronger in studies that were published before 1990 compare the diseases with the largest univariate effect sizes
(gP ¼ .47) than in later studies (gP ¼ .19–.25). (asthma, chronic fatigue syndrome, epilepsy, kidney/liver
Furthermore, the Qb-statistics indicates that levels of disease, or migraine/tension-type headache) against the
externalizing, but not of internalizing and total problems, other chronic conditions, a dummy variable was created
vary by the target of comparison. Significantly stronger for the type of illness. As shown in Table IV, all univariate
between-group differences in externalizing problems were moderator effects remained significant.
observed in studies that used a healthy control group
(gP ¼ .30) than in studies that compared behavior problems
of children with chronic illnesses to test norms (gP ¼ .20). Discussion
The effect size of the latter group was heterogeneous. The present study found elevated levels of behavior prob-
Finally, effect sizes for externalizing problems varied lems in children and adolescents with chronic physical
by the sociodemographic equivalence of the patient group illnesses. On average, elevations of internalizing problems,
and control group. The nonoverlap of the CI indicates that social problems, attention problems, and thought prob-
significantly weaker effect sizes were observed in studies lems were larger than elevations of externalizing problems.
with nonequivalent control groups (gP ¼ .10) than in other The effect sizes varied between the types of illnesses as well
studies (gP ¼ .20–.28). However, the effect sizes did not as raters of behavior problems, country and, in part, by age,
vary by the representativeness of the samples. Most effect gender, year of publication, study design, and equivalence
sizes of the subgroups were homogeneous. Nonetheless, of patient and control group. The discussion will be orga-
the test for heterogeneity was significant for the levels of nized according to the order of the research questions.
total problems in random samples.
We also ran all analyses with the modified total prob- Levels of Internalizing, Externalizing, and
lems and internalizing problems scale (Appendix A3). The Total Problems
results were very similar to the findings with the original We observed stronger elevations of internalizing problems
scales (e.g., mean difference for total problems gp ¼ .51, than of externalizing problems of children and adolescents
internalizing problems: gp ¼ .53). Only the moderating with chronic physical illness. This difference cannot be ex-
effect of age on the level of internalizing problems was plained by the fact that the internalizing score includes so-
not replicated in that analysis. matic complaints which may reflect illness-specific somatic
Because some moderator variables may be correlated, symptoms (Perrin et al., 1991). First, our meta-analysis
we concluded our investigation by testing whether the uni- showed that the main results remain unchanged when
variate moderator effects could be replicated in multivariate excluding the somatic complaints subscale. Note that
analysis. Only those moderators that were significant in at other correcting procedures of individual studies, such
least one univariate analysis were included. In order to as deleting other ambiguous items, led to similar results
Table IV. Multivariate Analysis of Moderating Effects of Study Characteristics (Multiple Linear Regression Analysis)
B b Z B b Z B b Z
Kind of illness (2 ¼ asthma/CFS/epilepsy/migraine/ .17 .22 10.11*** .25 .26 6.46*** .14 .15 3.70***
tension-type headache, 1 ¼ others)
Mean age (1 ¼ lower than 6 years, 2 ¼ others) .36 .22 10.88*** .31 .18 4.53*** .17 .11 2.64**
Percentage girls (2 ¼ larger 33%, 1 ¼ others) .11 .09 4.28*** .09 .08 1.99* .16 .15 3.76***
Country (2 ¼ developed country, 1 ¼ developing/ .04 .05 2.56* .11 .13 3.35*** .13 .17 4.13***
threshold country)
Year of publication (1 ¼ before 1990, 2 ¼ others) .13 .08 3.80*** .04 .02 0.58 .23 .17 4.13***
Target of comparison (2 ¼ test norms, 1 ¼ healthy .02 .03 1.22 .00 .00 0.07 .09 .11 2.55*
control group)
Equivalence of patients and control condition .07 .03 1.37 .14 .06 1.53 .20 .10 2.32*
(1 ¼ no, 2 ¼ yes/not tested)
(Constant) .04 0.29 .16 0.56 .57 2.14*
R2 .12 .13 .14
n 493 548 528
Note. *p < .05, **p < .01, ***p < .001.
Behavior Problems and Chronic Illness 1013
(e.g., Gleissner et al., 2008). Second, similar differences are physical illnesses whereas elevations of externalizing prob-
found when comparing levels of anxiety/depression lems are rather illness-specific. Most illnesses with the
and withdrawn behavior with delinquent and aggressive strongest elevations of externalizing problems, i.e. epilepsy,
behavior. The difference indicates that sources of internal- migraine/tension-type headache, and hearing impairment,
izing problems (e.g., restrictions of positive activities) may are always or often associated with impaired brain func-
be more widespread and/or may have stronger effects on tion. Neuroimaging studies have identified frontal lobe
children with chronic physical illness than sources of brain abnormalities in patients with epilepsy and migraine
externalizing problems. (e.g., Herrman et al., 2002; Schmitz et al., 2010). These
The analysis of the CBCL subscales indicated that, in abnormalities are associated with executive function defi-
addition to somatic problems, children with chronic illness cits, which could affect the inhibition of externalizing prob-
are particularly at increased risk for attention and social lems. Side effects of medications, such as corticosteroids,
problems. Attention problems are, in part, consequences may in part explain elevated externalizing problems among
of physical diseases, as in the case of epilepsy and chronic children with kidney and liver diseases (Soliday, Grey, &
headache (Hernandez, Sauerwein, Jambaqué, de Guise, Lande, 1999). The observed lower level of externalizing
Lussier, & Lortie, 2003). In addition, thinking or worrying problems in children with cleft lip and palate as compared
about ones’ illness may impair their attention. Social prob- to healthy controls or test norms are difficult to interpret
lems refer to being too dependent upon others, not getting because the subscales of delinquent and aggressive behav-
along with peers or getting teased (Achenbach, 1991). iors showed elevated scores. More research is needed before
Elevations of these symptoms indicate that chronic ill- definite conclusions can be drawn.
nesses often impair peer relations and that they may also Differences between illnesses in the levels of inter-
impair the development of autonomy. nalizing problems were impressive. The largest effect size
The present meta-analysis showed that adolescents do was observed for chronic fatigue syndrome which may, in
not only report fewer behavior problems than their parents, part, be based on a symptom overlap of tiredness, unspe-
but also that similar differences exist when comparing cific pain, sleep problems, and other somatic symptoms.
teacher and adolescent reports. As externalizing problems However, large elevations of internalizing problems per-
can often be easily observed by parents or teachers, adult sisted after eliminating the somatic complaints subscale
reports on externalizing behaviors may, therefore, be re- (Appendix 3). Fatigue is also likely to interfere with ma-
garded as quite objective (Frick, Barry, & Kamphaus, ny positive aspects of daily life which may cause internaliz-
2009). Our results may indicate that children with chronic ing problems, such as withdrawal or depressed symptoms.
illness tend to underreport their symptoms, for example, Interestingly, young people with chronic fatigue syndrome
because they want to present themselves as healthy func- did not show elevated levels of externalizing problems,
tioning individuals (Huberty, Austin, Harezlak, Dunn, & possibly because being tired inhibits such behaviors.
Ambrosius, 2000). Alternatively, parents and teachers may While children with chronic fatigue syndrome had
underestimate the ability of the children to adapt toward specific behavior problems, young people with migraine/
their illness. For example, caring for a child with chronic chronic-tension type headache and epilepsy showed above-
illness causes parental distress (e.g., Ashkani, Dehbozorgi, average levels of externalizing and internalizing problems.
& Tahamtan, 2004) which could lead to biased estimations Above-average levels of internalizing problems in these ill-
of child problems (Youngstrom, Loeber, & Stouthamer- nesses may be based on restrictions of positive activities,
Loeber, 2000). lack of perceived control over symptoms, and changes of
Because the YSR—in contrast to the CBCL—was only the brain. For example, epileptic activity in certain areas of
used with 11-year-olds and older adolescents, differences the brain directly causes paroxysmal anxiety (Beyenburg,
between studies with the CBCL/TRF and the YSR might Mitchell, Schmidt, Elger, & Reuber, 2005).
also reflect different age ranges studied. However, as the Although the effect sizes of most diseases were homo-
total, internalizing and externalizing CBCL scores of 11-year- geneous, they varied in particular between studies on
olds and older adolescents were significantly higher than the kidney/liver disease. This heterogeneity is probably based
YSR scores, the diverging age ranges of both instruments on the fact that the specific illnesses within this group were
cannot explain the observed rater differences. quite heterogeneous, for example, when needing organ
transplantation or showing a less severe disease.
Comparison of Illnesses In addition to the analysis of externalizing and inter-
Our results indicate that elevated levels of internalizing nalizing problems, our study also analyzed problems in the
problems are observed in children with most chronic fields of attention, thought, and social relations. Changes
1014 Pinquart and Shen
in brain function could explain the observed high levels of with chronic illness with a healthy control group (gp ¼ .60
attention problems in children with epilepsy and spina vs. gp ¼ .21/gp ¼ .09, Q(2,327) ¼ 53.11, p < .001).
bifida. For example, Hernandez et al. (2003) observed We only found weak evidence for the suggested higher
that especially abnormities of the frontal lobe of patients effect sizes in studies that compared children with chronic
with epilepsy were associated with attention problems. illnesses to healthy peers rather than test norms. Probably
Changes in the brain and the associated externalizing prob- children with severe chronic physical illnesses were under-
lems may also contribute to the high levels of social prob- represented in the normative samples of the CBCL so that
lems that were found in children with epilepsy. Observed similarities between the normative samples and healthy
high levels of social problems in children with spina bifida control groups prevail.
may be affected by the high levels of social withdrawal
which inhibits the development of peer-relations. Strengths and Limitations
Compared to previous meta-analyses on behavior problems
Moderator Effects of Sociodemographic with chronic illness, we were able to include a much larger
Characteristics data set. In order to collect a representative sample of
While in a meta-analysis on congenital heart disease studies, we also did not limit our analysis to studies pub-
Karsdorp et al. (2007) had found lower levels of behavior lished in English. This allowed for comparing a much larger
problems in children younger than 6 years than in those number of chronic illnesses and for assessing a broader
older than 10 years, our study found that the youngest range of moderator variables.
group also differed from 6 to 10-year-olds. Below-average Nonetheless, some limitations have to be mentioned.
levels of behavior problems in younger children with First, the available cross-sectional data did not allow for
chronic illness may indicate that some negative psychoso- testing whether behavior problems were the consequences
cial consequences of their illness do not yet appear that of the physical illness or whether some of these problems
early, such as being bullied by class mates. developed independently or even before the diagnosis of a
Interestingly, we observed larger effect sizes for total chronic physical disease. However, we were able to show
and externalizing problems in studies with higher percent- that children with chronic physical illness are at increased
ages of boys, whereas levels of internalizing problems did risk for these problems. Second, only a few studies were
not increase significantly in the case of a higher percentage available for some kinds of illnesses, such as HIV-infection/
of girls. Thus, our results indicate that boys are more likely AIDS and visual impairment. Third, due to space limita-
to react toward their illness with externalizing symptoms, tions, we did not include data on social competence that
possibly because this behavior is more consistent with the are also assessed with the CBCL. Fourth, we limited our
male gender role (Steinberg, 2008). Studies not specific to analysis to the CBCL scales that have most often been used
chronic physical illnesses have shown that gender differ- in studies on chronic illness. Too few studies were available
ences in externalizing problems emerge, on average, earlier for the DSM-oriented CBCL scales that were introduced by
than gender differences in internalizing problems (Steinberg, Achenbach, Dumenici, and Rescorla (2003). Fifth, in order
2008). This could lead to the observed lack of moderating to address the critique by Perrin et al. (1991) on the use of
effect of gender on the level of internalizing symptoms. the somatic complaints subscale with people with chronic
illness, we had to use a crude estimation of the corrected
Characteristics of the Publication internalizing and total problems. Finally, we did not ana-
The expected larger effect sizes of older studies were only lyze whether similar results would be found when using
observed with regard to externalizing behavior. This result other screening measures for behavior problems.
may indicate that more progress has been made in the
prevention or treatment of externalizing rather than inter-
nalizing problems. In fact, in the last two decades more Conclusions
studies on the prevention of externalizing problems have Despite these limitations, several conclusions can be drawn
been published than on the prevention of internalizing from this meta-analysis. First, we conclude that elevated
problems (e.g., Röhrle, 2007). Alternatively, the effect of levels of internalizing and total problems are observed in
time of measurement might also be based on the use of almost all chronic illnesses. However, the level and profile
different test norms. However, a supplementary analysis of behavior problems (e.g., whether internalizing and/or
showed that larger elevations of externalizing problems externalizing problems are elevated) varies between dis-
are also found in older studies that compared children eases. Second, the present meta-analysis highlights that
Behavior Problems and Chronic Illness 1015
elevated levels of problems are also found in other areas, approaches to constructing scales from the same
such as attention or social problems, whereas many studies item pools. Journal of Clinical Child Psychology and
focus exclusively on the internalizing and externalizing Psychiatry, 32, 328–340.
scores of the CBCL. Third, the inclusion of the somatic Ashkani, H., Dehbozorgi, D. R., & Tahamtan, A. (2004).
complaints subscale is unlikely to cause a serious bias when Depression among parents of children with chronic
working with total problem scores and internalizing prob- and disabling disease. Iran Journal of Medical Science,
lem scores of children with chronic illness. Nonetheless, 29, 90–93.
we agree with Perrin et al. (1991) that the results from Barlow, J. H., & Ellard, D. R. (2006). The psychosocial
somatic complaints scale would be difficult to interprete well-being of children with chronic disease, their
in this group of children. Fourth, due to the systematic parents and siblings: An overview of the research
differences between results from adolescent self-reports evidence base. Child: Care, Health and Development,
and parent/teacher reports, we recommend the use of dif- 32, 19–31.
ferent sources of information. If researchers or clinicians do Bennett, D. S. (1994). Depression among children
not want to overlook patients with elevated levels of behav- with chronic medical problems: A meta-analysis.
ior problems, they should not rely exclusively on patient Journal of Pediatric Psychology, 19, 149–169.
self-reports. Fifth, we conclude that boys and girls with Beyenburg, S., Mitchell, A. J., Schmidt, D., Elger, C. E.,
chronic physical illness should routinely be screened for & Reuber, M. (2005). Anxiety in patients with epi-
internalizing problems whereas boys would be the main lepsy: Systematic review and suggestions for clinical
target for screening for externalizing problems, given the management. Epilepsy and Behavior, 7, 161–171.
observed gender differences. Sixth, the observed elevated Bleyer, W. A. (2002). Cancer in older adolescents and
levels of behavior problems call for a multidisciplinary team young adults. Medical and Pediatric Oncology, 38,
approach wherein medical and psychosocial personnel es- 1–10.
tablish a collaborative treatment approach. In addition to Cohen, J. (1988). Statistical power analysis for the
treating the somatic disease, psychosocial interventions are behavioral sciences. Hillsdale: Erlbaum.
needed for preventing and treating behavior problems. For Frick, P. J., Barry, C. T., & Kamphaus, R. W. (2009).
example, Perrin, MacLean, Gortmarker, and Asher (1992) Clinical assessment of child and adolescent personality
have shown that a stress management intervention reduced and behavior (3rd ed.). New York: Springer.
the level of behavior problems of young asthma patients. Gleissner, U., Fritz, N. E., Von Lehe, M., Sassen, R.,
Seventh, given the higher levels of behavior problems Elger, C. E., & Helmstaegdter, C. (2008). The validi-
among children with chronic physical illnesses from devel- ty of the Child Behavior Checklist for children with
oping/threshold countries, more psychosocial services epilepsy. Epilepsy & Behavior, 12, 276–280.
aimed at preventing or treating these problems are needed Hedges, L. V., & Pigott, T. D. (2001). The power of
in these countries. Finally, future research should analyze statistical tests in meta-analysis. Psychological
the interplay of processes that mediate the effect of chronic Methods, 6, 203–217.
physical illness on different forms of behavior problems. Herrman, B., Seidenberg, M., Bell, B., Rutecki, P.,
Such studies would be helpful for identifying starting Sheth, R., Ruggles, K., . . . Magnotta, V. (2002).
points of interventions. The neurodevelopmental impact of childhood-onset
temporal lobe epilepsy on brain structure and
function. Epilepsia, 43, 1062–1071.
Supplementary Data Hernandez, M.-T., Sauerwein, H. C., Jambaqué, I., de
Supplementary data available at JPEPSY online. Guise, E., Lussier, F., & Lortie, A. (2003). Attention,
memory, and behavioral adjustment in children with
Conflicts of interest: None declared. frontal lobe epilepsy. Epilepsy & Behavior, 4,
522–536.
Huberty, T. J., Austin, J. K., Harezlak, J., Dunn, D. W.,
References & Ambrosius, W. T. (2000). Informant agreement in
Achenbach, T. M. (1991). Integrative guide for the 1991 behavior ratings for children with epilepsy. Epilepsy
CBCL/4 18, YSR, and TRF profiles. Burlington: and Behavior, 1, 427–435.
Department of Psychiatry, University of Vermont. Karsdorp, P. A., Everaerd, W., King, T. M., & Mulder, B.
Achenbach, T. M., Dumenici, L., & Rescorla, L. A. J. (2007). Psychological and cognitive functioning
(2003). DSM-oriented and empirically based in children and adolescents with congenital heart
1016 Pinquart and Shen
disease: A meta-analysis. Journal of Pediatric Rodenburg, R., Stams, G. J., Meijer, A. M.,
Psychology, 32, 527–541. Aldenkamp, A. P., & Dekovic, M. (2005).
Lavigne, J. V., & Faier-Routman, J. (1992). Psychological Psychopathology in children with epilepsy: A
adjustment to pediatric physical disorders: A meta- meta-analysis. Journal of Pediatric Psychology, 30,
analysis. Journal of Pediatric Psychology, 17, 133–157. 453–468.
LeBovidge, J. S, Lavigne, J. V, Donenberg, G. R., & Röhrle, B. (2007). Prävention und Gesundheitsförderung:
Miller, M. L. (2005). Adjustment to chronic arthritis Kinder und Jugendliche [Prevention and health
of childhood: The roles of illness-related stress and promotion: Children and adolescencts]. Tübingen: dgvt.
attitude toward illness. Journal of Pediatric Psychology, Rosenthal, T. (1991). Meta-analytic procedures for social
30, 273–268. research. Newbury Park: Sage.
Lipsey, M. W., & Wilson, D. B. (2001). Practical Schmitz, N., Arkink, E. B., Mulder, M., Rubia, K.,
meta-analysis. Thousand Oaks, CA: Sage. Admiraal-Behloul, F., Schoonmann, G. G., . . .
McQuaid, E. L., Kopel, S. J., & Nassau, J. H. (2001). van Buchem, M. A. (2008). Frontal lobe structure
Behavioral adjustment in children with asthma: A and executive function in migraine patients.
meta-analysis. Journal of Developmental & Behavioral Neuroscience Letters, 440, 92–96.
Pediatrics, 22, 430–439. Soliday, E., Grey, S., & Lande, M. B. (1999). Behavioral
Perrin, E. C., Stein, R. E., & Drotar, D. (1991). Cautions effects of corticosteroids in steroid-sensitive nephrotic
in using the Child Behavior Checklist: Observations syndrome. Pediatrics, 104, 1–4.
based on research about children with a chronic Steinberg, L. (2008). Adolescence (8th ed.). Boston:
illness. Journal of Pediatric Psychology, 16, 411–421. McGraw-Hill.
Perrin, J. M., MacLean, W. E., Gortmarker, S. L., & van der Lee, J., Mokkink, L. B., Grootenhuis, M. A.,
Asher, K. N. (1992). Improving the psychological Heymans, H. S., & Offringa, M. (2007). Definitions
status of children with asthma: A randomized con- and measurement of chronic health conditions in
trolled trial. Developmental and Behavioral Pediatrics, childhood: A systematic review. JAMA, 29,
13, 241–247. 2741–2751.
Pinquart, M., & Shen, Y. (2011). Depressive symptoms World Bank. (2010). Country and Lending Groups.
in children and adolescents with chronic physical ill- Retrieved from http://data.worldbank.org/about/country-
ness: An updated meta-analysis. Journal of Pediatric classifications/country-and-lending-groups.
Psychology, 36, 375–384. Youngstrom, E., Loeber, R., & Stouthamer-Loeber, M.
Reijntjes, A., Kamphuis, J. H., Prinzie, P., Boelen, P. A., (2000). Patterns and correlates of agreement between
van der Schoot, M., & Telch, M. J. (2010). parent, teacher, and male adolescent ratings of
Prospective linkages between peer victimization and externalizing and internalizing problems.
externalizing problems in children: A meta-analysis. Journal of Consulting and Clinical Psychology, 68,
Aggressive Behavior, 35, 1–8. 1038–1050.
Copyright of Journal of Pediatric Psychology is the property of Oxford University Press / UK and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express
written permission. However, users may print, download, or email articles for individual use.